Rotational scanning techniques for hyperspectral imaging

Timothy Kelman, Stephen Marshall, Jinchang Ren, John Gilchrist

Research output: Contribution to conferencePaper

Abstract

Since hyperspectral images form three dimensional data structures, data acquisition usually requires that two of the dimensions remain constant, whilst the third is varied. A popular method for capturing hyperspectral data is pushbroom scanning. This technique builds an image by incrementally capturing all available wavelengths on a spatial line scan over time. One drawback of this technique is that an accurate way of linearly moving the camera or the object is required. Rotational movement is often easier to achieve or naturally available, hence the work presented here investigates an alternative approach to the traditional pushbroom method. If the rotation of the object is perfectly aligned (i.e. the axis of rotation is coincident with the leftmost pixel for anti-clockwise rotation) the acquired hypercube is, in fact, a polar representation of the scanned object. This ideal polar representation can be easily converted to Cartesian form using existing toolboxes [1] to produce a conventional hypercube.
While rotational movement is often easier to achieve than the translational motion required by pushbroom, near perfect rotation is often not possible. There are two main sources of error when using rotational scanning: horizontal offset and vertical offset. These offsets occur when the axis of rotation does not precisely coincide with the desired pixel as described above. This paper explains, using examples, a method to compute these two parameters such that they can be used to compensate for the errors caused by these offsets. Although polar to Cartesian conversion can be calculated, there is no guarantee that each pixel stored in polar form can be mapped to a unique point in Cartesian space. As such, a suitable interpolation method is required to compensate for this. A number of such techniques are discussed and evaluated in this paper.
This work will lead to the development of a tool which will be capable of automatically estimating the offset parameters. Further investigation into other forms of error in rotational scanning will also be carried out (e.g. non-circular rotation).
LanguageEnglish
Pages76-85
Number of pages10
Publication statusPublished - 15 May 2012
EventHyperspectral Imaging Conference - Rome, Italy
Duration: 15 May 201216 May 2012

Conference

ConferenceHyperspectral Imaging Conference
CountryItaly
CityRome
Period15/05/1216/05/12

Fingerprint

Scanning
Pixels
Data structures
Hyperspectral imaging
Data acquisition
Interpolation
Cameras
Wavelength

Keywords

  • hyperspectral imaging
  • rotational scanning

Cite this

Kelman, T., Marshall, S., Ren, J., & Gilchrist, J. (2012). Rotational scanning techniques for hyperspectral imaging. 76-85. Paper presented at Hyperspectral Imaging Conference, Rome, Italy.
Kelman, Timothy ; Marshall, Stephen ; Ren, Jinchang ; Gilchrist, John . / Rotational scanning techniques for hyperspectral imaging. Paper presented at Hyperspectral Imaging Conference, Rome, Italy.10 p.
@conference{fd197e546c904f2db98c81b6df06e45d,
title = "Rotational scanning techniques for hyperspectral imaging",
abstract = "Since hyperspectral images form three dimensional data structures, data acquisition usually requires that two of the dimensions remain constant, whilst the third is varied. A popular method for capturing hyperspectral data is pushbroom scanning. This technique builds an image by incrementally capturing all available wavelengths on a spatial line scan over time. One drawback of this technique is that an accurate way of linearly moving the camera or the object is required. Rotational movement is often easier to achieve or naturally available, hence the work presented here investigates an alternative approach to the traditional pushbroom method. If the rotation of the object is perfectly aligned (i.e. the axis of rotation is coincident with the leftmost pixel for anti-clockwise rotation) the acquired hypercube is, in fact, a polar representation of the scanned object. This ideal polar representation can be easily converted to Cartesian form using existing toolboxes [1] to produce a conventional hypercube.While rotational movement is often easier to achieve than the translational motion required by pushbroom, near perfect rotation is often not possible. There are two main sources of error when using rotational scanning: horizontal offset and vertical offset. These offsets occur when the axis of rotation does not precisely coincide with the desired pixel as described above. This paper explains, using examples, a method to compute these two parameters such that they can be used to compensate for the errors caused by these offsets. Although polar to Cartesian conversion can be calculated, there is no guarantee that each pixel stored in polar form can be mapped to a unique point in Cartesian space. As such, a suitable interpolation method is required to compensate for this. A number of such techniques are discussed and evaluated in this paper.This work will lead to the development of a tool which will be capable of automatically estimating the offset parameters. Further investigation into other forms of error in rotational scanning will also be carried out (e.g. non-circular rotation).",
keywords = "hyperspectral imaging , rotational scanning",
author = "Timothy Kelman and Stephen Marshall and Jinchang Ren and John Gilchrist",
year = "2012",
month = "5",
day = "15",
language = "English",
pages = "76--85",
note = "Hyperspectral Imaging Conference ; Conference date: 15-05-2012 Through 16-05-2012",

}

Kelman, T, Marshall, S, Ren, J & Gilchrist, J 2012, 'Rotational scanning techniques for hyperspectral imaging' Paper presented at Hyperspectral Imaging Conference, Rome, Italy, 15/05/12 - 16/05/12, pp. 76-85.

Rotational scanning techniques for hyperspectral imaging. / Kelman, Timothy; Marshall, Stephen; Ren, Jinchang; Gilchrist, John .

2012. 76-85 Paper presented at Hyperspectral Imaging Conference, Rome, Italy.

Research output: Contribution to conferencePaper

TY - CONF

T1 - Rotational scanning techniques for hyperspectral imaging

AU - Kelman, Timothy

AU - Marshall, Stephen

AU - Ren, Jinchang

AU - Gilchrist, John

PY - 2012/5/15

Y1 - 2012/5/15

N2 - Since hyperspectral images form three dimensional data structures, data acquisition usually requires that two of the dimensions remain constant, whilst the third is varied. A popular method for capturing hyperspectral data is pushbroom scanning. This technique builds an image by incrementally capturing all available wavelengths on a spatial line scan over time. One drawback of this technique is that an accurate way of linearly moving the camera or the object is required. Rotational movement is often easier to achieve or naturally available, hence the work presented here investigates an alternative approach to the traditional pushbroom method. If the rotation of the object is perfectly aligned (i.e. the axis of rotation is coincident with the leftmost pixel for anti-clockwise rotation) the acquired hypercube is, in fact, a polar representation of the scanned object. This ideal polar representation can be easily converted to Cartesian form using existing toolboxes [1] to produce a conventional hypercube.While rotational movement is often easier to achieve than the translational motion required by pushbroom, near perfect rotation is often not possible. There are two main sources of error when using rotational scanning: horizontal offset and vertical offset. These offsets occur when the axis of rotation does not precisely coincide with the desired pixel as described above. This paper explains, using examples, a method to compute these two parameters such that they can be used to compensate for the errors caused by these offsets. Although polar to Cartesian conversion can be calculated, there is no guarantee that each pixel stored in polar form can be mapped to a unique point in Cartesian space. As such, a suitable interpolation method is required to compensate for this. A number of such techniques are discussed and evaluated in this paper.This work will lead to the development of a tool which will be capable of automatically estimating the offset parameters. Further investigation into other forms of error in rotational scanning will also be carried out (e.g. non-circular rotation).

AB - Since hyperspectral images form three dimensional data structures, data acquisition usually requires that two of the dimensions remain constant, whilst the third is varied. A popular method for capturing hyperspectral data is pushbroom scanning. This technique builds an image by incrementally capturing all available wavelengths on a spatial line scan over time. One drawback of this technique is that an accurate way of linearly moving the camera or the object is required. Rotational movement is often easier to achieve or naturally available, hence the work presented here investigates an alternative approach to the traditional pushbroom method. If the rotation of the object is perfectly aligned (i.e. the axis of rotation is coincident with the leftmost pixel for anti-clockwise rotation) the acquired hypercube is, in fact, a polar representation of the scanned object. This ideal polar representation can be easily converted to Cartesian form using existing toolboxes [1] to produce a conventional hypercube.While rotational movement is often easier to achieve than the translational motion required by pushbroom, near perfect rotation is often not possible. There are two main sources of error when using rotational scanning: horizontal offset and vertical offset. These offsets occur when the axis of rotation does not precisely coincide with the desired pixel as described above. This paper explains, using examples, a method to compute these two parameters such that they can be used to compensate for the errors caused by these offsets. Although polar to Cartesian conversion can be calculated, there is no guarantee that each pixel stored in polar form can be mapped to a unique point in Cartesian space. As such, a suitable interpolation method is required to compensate for this. A number of such techniques are discussed and evaluated in this paper.This work will lead to the development of a tool which will be capable of automatically estimating the offset parameters. Further investigation into other forms of error in rotational scanning will also be carried out (e.g. non-circular rotation).

KW - hyperspectral imaging

KW - rotational scanning

UR - http://istituto.ingv.it/l-ingv/convegni-e-seminari/archivio-congressi/convegni-2012/hsi2012-1/hsi2012/view

M3 - Paper

SP - 76

EP - 85

ER -

Kelman T, Marshall S, Ren J, Gilchrist J. Rotational scanning techniques for hyperspectral imaging. 2012. Paper presented at Hyperspectral Imaging Conference, Rome, Italy.